I’ve found myself shaking my head in disbelief over the past weeks as stories about incredible uptake rates for Version 1803 have appeared. Usually credible writers such as Peter Bright at Ars Technica and Liam Tung at ZDNet have been taken in on some numbers based on observations from AdDuplex which describes itself as a “leading cross-promotion network for Windows Phone and Windows applications.” Alas, as my old buddy and fellow Windows Insider MVP Ed Bott observes, “the inclusion of Windows Phone, a now defunct product … should be a red flag” when it comes to assessing AdDuplex’s credibility and veracity as a broad base for statistical anything. In fact, Mr. Bott convincingly argues at ZDNet that characterizing AdDuplex-based Win10 uptake projections mostly bogus is absolutely correct.
It’s impossible for half the user base to have updated to 1803 already.
Why Say AdDuplex-based Win10 Uptake Projections Mostly Bogus?
AdDuplex’s numbers extrapolated to the entire Windows installed base indicate that half that population has already updated to Version 1803. As Mr. Bott observes in his article (and I had independently concluded on my own) this is flatly impossible. Microsoft has just reported on June 1 at its Build 2018 meet-up, as per this Neowin story, that “there are over 700 million active Windows 10 devices.” Thus, that would mean somewhere in the neighborhood of 350 million devices already upgraded to 1803. Simply put, that’s not true (or possible).
Here’s why. The extrapolation comes from data collected on about five thousand Windows Store apps with the AdDuplex SDK embedded inside them. Bott reports a principal at AdDuplex puts the number of devices sampled at “over 100,000.” As Bott observes, that’s less than 2.14 x 10-6 (0.00000214) of the total Windows population. (I adjusted his 600M baseline to the 700M baseline confirmed by MS this morning.) The kicker is that only AdDuplex apps use the SDK. Many of those apps are mobile (a mostly defunct and vanishing platform). In fact, they focus primarily on what Bott describes as “casual games … and some utility apps.”
A Shaky Foundation for All Windows Users
What kind of basis is that for making projections about the Windows user base? Bott submits, and I concur, that it’s not a very good one. For one thing, it completely omits the 200M enterprise/business users of Windows 10 (as reported in this 5/8 ComputerWorld story), none of whom are likely to have anything to do with AdDuplex data acquisition. For another, the mobile focus in AdDuplex represents an odd and vanishing subset of Windows 10 users. And finally, there’s the practical observation that MS deliberately stages its release of new feature upgrades, and withholds the upgrade through Windows Update for older and potentially more problematic PCs. It also blocks the update for machines with known gotchas (such as the Intel 600p series SSDs, which didn’t work with 1803 until KB4100403 was released on May 23) until fixes are ready.
Flatly stated, there’s no way in heaven or hell that half of the user base is at 1803 just yet. It’s simply not possible, and thus, completely unlikely as well. As the old data analysis maxim goes: “Garbage in, garbage out!”